1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
5 #ifndef LLITE_INTERNAL_H
6 #define LLITE_INTERNAL_H
8 #include <linux/ext2_fs.h>
9 #ifdef CONFIG_FS_POSIX_ACL
10 # include <linux/fs.h>
12 # include <linux/xattr_acl.h>
14 #ifdef HAVE_LINUX_POSIX_ACL_XATTR_H
15 # include <linux/posix_acl_xattr.h>
19 #include <lustre_debug.h>
20 #include <lustre_ver.h>
21 #include <linux/lustre_version.h>
22 #include <lustre_disk.h> /* for s2sbi */
25 struct lustre_intent_data {
26 __u64 it_lock_handle[2];
32 /* If there is no FMODE_EXEC defined, make it to match nothing */
37 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
38 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
40 #ifdef HAVE_VFS_INTENT_PATCHES
41 static inline struct lookup_intent *ll_nd2it(struct nameidata *nd)
43 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
51 struct ll_dentry_data {
54 struct obd_client_handle lld_cwd_och;
55 struct obd_client_handle lld_mnt_och;
56 #ifndef HAVE_VFS_INTENT_PATCHES
57 struct lookup_intent *lld_it;
59 unsigned int lld_sa_generation;
60 cfs_waitq_t lld_waitq;
63 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
65 extern struct file_operations ll_pgcache_seq_fops;
67 #define LLI_INODE_MAGIC 0x111d0de5
68 #define LLI_INODE_DEAD 0xdeadd00d
69 #define LLI_F_HAVE_OST_SIZE_LOCK 0
70 #define LLI_F_HAVE_MDS_SIZE_LOCK 1
71 #define LLI_F_CONTENDED 2
72 #define LLI_F_SRVLOCK 3
74 struct ll_inode_info {
76 struct semaphore lli_size_sem; /* protect open and change size */
77 void *lli_size_sem_owner;
78 struct semaphore lli_write_sem;
79 struct lov_stripe_md *lli_smd;
80 char *lli_symlink_name;
83 unsigned long lli_flags;
84 cfs_time_t lli_contention_time;
86 /* this lock protects s_d_w and p_w_ll and mmap_cnt */
88 #ifdef HAVE_CLOSE_THREAD
89 struct list_head lli_pending_write_llaps;
90 struct list_head lli_close_item;
91 int lli_send_done_writing;
93 atomic_t lli_mmap_cnt;
95 /* for writepage() only to communicate to fsync */
98 struct posix_acl *lli_posix_acl;
100 struct list_head lli_dead_list;
102 struct semaphore lli_och_sem; /* Protects access to och pointers
103 and their usage counters */
104 /* We need all three because every inode may be opened in different
106 struct obd_client_handle *lli_mds_read_och;
107 __u64 lli_open_fd_read_count;
108 struct obd_client_handle *lli_mds_write_och;
109 __u64 lli_open_fd_write_count;
110 struct obd_client_handle *lli_mds_exec_och;
111 __u64 lli_open_fd_exec_count;
113 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
114 struct inode lli_vfs_inode;
117 /* metadata stat-ahead */
119 * "opendir_pid" is the token when lookup/revalid -- I am the owner of
122 pid_t lli_opendir_pid;
124 * since parent-child threads can share the same @file struct,
125 * "opendir_key" is the token when dir close for case of parent exit
126 * before child -- it is me should cleanup the dir readahead. */
127 void *lli_opendir_key;
128 struct ll_statahead_info *lli_sai;
132 * Locking to guarantee consistency of non-atomic updates to long long i_size,
133 * consistency between file size and KMS, and consistency within
134 * ->lli_smd->lsm_oinfo[]'s.
136 * Implemented by ->lli_size_sem and ->lsm_sem, nested in that order.
139 void ll_inode_size_lock(struct inode *inode, int lock_lsm);
140 void ll_inode_size_unlock(struct inode *inode, int unlock_lsm);
142 // FIXME: replace the name of this with LL_I to conform to kernel stuff
143 // static inline struct ll_inode_info *LL_I(struct inode *inode)
144 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
146 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
147 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
149 CLASSERT(sizeof(inode->u) >= sizeof(struct ll_inode_info));
150 return (struct ll_inode_info *)&(inode->u.generic_ip);
154 /* default to about 40meg of readahead on a given system. That much tied
155 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
156 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
158 /* default to read-ahead full files smaller than 2MB on the second read */
159 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
164 RA_STAT_DISTANT_READPAGE,
165 RA_STAT_MISS_IN_WINDOW,
166 RA_STAT_FAILED_GRAB_PAGE,
167 RA_STAT_FAILED_MATCH,
172 RA_STAT_MAX_IN_FLIGHT,
173 RA_STAT_WRONG_GRAB_PAGE,
178 unsigned long ra_cur_pages;
179 unsigned long ra_max_pages;
180 unsigned long ra_max_read_ahead_whole_pages;
181 unsigned long ra_stats[_NR_RA_STAT];
184 /* LL_HIST_MAX=32 causes an overflow */
185 #define LL_HIST_MAX 28
186 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
187 #define LL_PROCESS_HIST_MAX 10
188 struct per_process_info {
190 struct obd_histogram pp_r_hist;
191 struct obd_histogram pp_w_hist;
194 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
195 struct ll_rw_extents_info {
196 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
199 #define LL_OFFSET_HIST_MAX 100
200 struct ll_rw_process_info {
203 loff_t rw_range_start;
205 loff_t rw_last_file_pos;
207 size_t rw_smallest_extent;
208 size_t rw_largest_extent;
209 struct file *rw_last_file;
213 enum stats_track_type {
214 STATS_TRACK_ALL = 0, /* track all processes */
215 STATS_TRACK_PID, /* track process with this pid */
216 STATS_TRACK_PPID, /* track processes with this ppid */
217 STATS_TRACK_GID, /* track processes with this gid */
221 /* flags for sbi->ll_flags */
222 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
223 #define LL_SBI_DATA_CHECKSUM 0x02 /* checksum each page on the wire */
224 #define LL_SBI_FLOCK 0x04
225 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
226 #define LL_SBI_ACL 0x10 /* support ACL */
227 #define LL_SBI_JOIN 0x20 /* support JOIN */
228 #define LL_SBI_LOCALFLOCK 0x40 /* Local flocks support by kernel */
229 #define LL_SBI_LRU_RESIZE 0x80 /* support lru resize */
230 #define LL_SBI_LLITE_CHECKSUM 0x100 /* checksum each page in memory */
232 /* default value for ll_sb_info->contention_time */
233 #define SBI_DEFAULT_CONTENTION_SECONDS 60
234 /* default value for lockless_truncate_enable */
235 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
238 struct list_head ll_list;
239 /* this protects pglist and ra_info. It isn't safe to
240 * grab from interrupt contexts */
242 spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
243 spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
244 struct obd_uuid ll_sb_uuid;
245 struct obd_export *ll_mdc_exp;
246 struct obd_export *ll_osc_exp;
247 struct proc_dir_entry *ll_proc_root;
248 obd_id ll_rootino; /* number of root inode */
251 struct list_head ll_conn_chain; /* per-conn chain of SBs */
252 struct lustre_client_ocd ll_lco;
254 struct list_head ll_orphan_dentry_list; /*please don't ask -p*/
255 struct ll_close_queue *ll_lcq;
257 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
259 unsigned long ll_async_page_max;
260 unsigned long ll_async_page_count;
261 unsigned long ll_pglist_gen;
262 struct list_head ll_pglist; /* all pages (llap_pglist_item) */
264 unsigned ll_contention_time; /* seconds */
265 unsigned ll_lockless_truncate_enable; /* true/false */
267 struct ll_ra_info ll_ra_info;
268 unsigned int ll_namelen;
269 struct file_operations *ll_fop;
271 #ifdef HAVE_EXPORT___IGET
272 struct list_head ll_deathrow; /* inodes to be destroyed (b1443) */
273 spinlock_t ll_deathrow_lock;
275 /* =0 - hold lock over whole read/write
276 * >0 - max. chunk to be read/written w/o lock re-acquiring */
277 unsigned long ll_max_rw_chunk;
280 struct ll_rw_extents_info ll_rw_extents_info;
281 int ll_extent_process_count;
282 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
283 unsigned int ll_offset_process_count;
284 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
285 unsigned int ll_rw_offset_entry_count;
286 enum stats_track_type ll_stats_track_type;
287 int ll_stats_track_id;
289 dev_t ll_sdev_orig; /* save s_dev before assign for
292 /* metadata stat-ahead */
293 unsigned int ll_sa_max; /* max statahead RPCs */
294 unsigned int ll_sa_wrong; /* statahead thread stopped for
296 unsigned int ll_sa_total; /* statahead thread started
298 unsigned long long ll_sa_blocked; /* ls count waiting for
300 unsigned long long ll_sa_cached; /* ls count got in cache */
301 unsigned long long ll_sa_hit; /* hit count */
302 unsigned long long ll_sa_miss; /* miss count */
305 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
310 struct task_struct *lrr_reader;
311 struct list_head lrr_linkage;
315 * per file-descriptor read-ahead data.
317 struct ll_readahead_state {
320 * index of the last page that read(2) needed and that wasn't in the
321 * cache. Used by ras_update() to detect seeks.
323 * XXX nikita: if access seeks into cached region, Lustre doesn't see
326 unsigned long ras_last_readpage;
328 * number of pages read after last read-ahead window reset. As window
329 * is reset on each seek, this is effectively a number of consecutive
330 * accesses. Maybe ->ras_accessed_in_window is better name.
332 * XXX nikita: window is also reset (by ras_update()) when Lustre
333 * believes that memory pressure evicts read-ahead pages. In that
334 * case, it probably doesn't make sense to expand window to
335 * PTLRPC_MAX_BRW_PAGES on the third access.
337 unsigned long ras_consecutive_pages;
339 * number of read requests after the last read-ahead window reset
340 * As window is reset on each seek, this is effectively the number
341 * on consecutive read request and is used to trigger read-ahead.
343 unsigned long ras_consecutive_requests;
345 * Parameters of current read-ahead window. Handled by
346 * ras_update(). On the initial access to the file or after a seek,
347 * window is reset to 0. After 3 consecutive accesses, window is
348 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
349 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
351 unsigned long ras_window_start, ras_window_len;
353 * Where next read-ahead should start at. This lies within read-ahead
354 * window. Read-ahead window is read in pieces rather than at once
355 * because: 1. lustre limits total number of pages under read-ahead by
356 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
357 * not covered by DLM lock.
359 unsigned long ras_next_readahead;
361 * Total number of ll_file_read requests issued, reads originating
362 * due to mmap are not counted in this total. This value is used to
363 * trigger full file read-ahead after multiple reads to a small file.
365 unsigned long ras_requests;
367 * Page index with respect to the current request, these value
368 * will not be accurate when dealing with reads issued via mmap.
370 unsigned long ras_request_index;
372 * list of struct ll_ra_read's one per read(2) call current in
373 * progress against this file descriptor. Used by read-ahead code,
374 * protected by ->ras_lock.
376 struct list_head ras_read_beads;
378 * The following 3 items are used for detecting the stride I/O
380 * In stride I/O mode,
381 * ...............|-----data-----|****gap*****|--------|******|....
382 * offset |-stride_pages-|-stride_gap-|
383 * ras_stride_offset = offset;
384 * ras_stride_length = stride_pages + stride_gap;
385 * ras_stride_pages = stride_pages;
386 * Note: all these three items are counted by pages.
388 unsigned long ras_stride_length;
389 unsigned long ras_stride_pages;
390 pgoff_t ras_stride_offset;
392 * number of consecutive stride request count, and it is similar as
393 * ras_consecutive_requests, but used for stride I/O mode.
394 * Note: only more than 2 consecutive stride request are detected,
395 * stride read-ahead will be enable
397 unsigned long ras_consecutive_stride_requests;
400 extern cfs_mem_cache_t *ll_file_data_slab;
401 struct lustre_handle;
402 struct ll_file_data {
403 struct ll_readahead_state fd_ras;
405 struct lustre_handle fd_cwlockh;
406 unsigned long fd_gid;
410 struct lov_stripe_md;
412 extern spinlock_t inode_lock;
414 extern struct proc_dir_entry *proc_lustre_fs_root;
416 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
418 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
419 return &lli->lli_vfs_inode;
421 return list_entry(lli, struct inode, u.generic_ip);
426 struct inode *icbd_parent;
427 struct dentry **icbd_childp;
431 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
433 #define LLAP_MAGIC 98764321
435 extern cfs_mem_cache_t *ll_async_page_slab;
436 extern size_t ll_async_page_slab_size;
437 struct ll_async_page {
439 /* only trust these if the page lock is providing exclusion */
440 unsigned int llap_write_queued:1,
441 llap_defer_uptodate:1,
446 llap_lockless_io_page:1;
448 struct page *llap_page;
449 struct list_head llap_pending_write;
450 struct list_head llap_pglist_item;
451 /* checksum for paranoid I/O debugging */
456 * enumeration of llap_from_page() call-sites. Used to export statistics in
457 * /proc/fs/lustre/llite/fsN/dump_page_cache.
460 LLAP_ORIGIN_UNKNOWN = 0,
461 LLAP_ORIGIN_READPAGE,
462 LLAP_ORIGIN_READAHEAD,
463 LLAP_ORIGIN_COMMIT_WRITE,
464 LLAP_ORIGIN_WRITEPAGE,
465 LLAP_ORIGIN_REMOVEPAGE,
466 LLAP_ORIGIN_LOCKLESS_IO,
469 extern char *llap_origins[];
471 #ifdef HAVE_REGISTER_CACHE
472 #define ll_register_cache(cache) register_cache(cache)
473 #define ll_unregister_cache(cache) unregister_cache(cache)
475 #define ll_register_cache(cache) do {} while (0)
476 #define ll_unregister_cache(cache) do {} while (0)
479 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
480 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
481 struct ll_ra_read *ll_ra_read_get(struct file *f);
483 /* llite/lproc_llite.c */
485 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
486 struct super_block *sb, char *osc, char *mdc);
487 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
488 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
489 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
491 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
492 struct super_block *sb, char *osc, char *mdc){return 0;}
493 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
494 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
495 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
497 memset(lvars, 0, sizeof(*lvars));
503 extern struct file_operations ll_dir_operations;
504 extern struct inode_operations ll_dir_inode_operations;
506 struct page *ll_get_dir_page(struct inode *dir, unsigned long n);
508 * p is at least 6 bytes before the end of page
510 typedef struct ext2_dir_entry_2 ext2_dirent;
512 static inline ext2_dirent *ext2_next_entry(ext2_dirent *p)
514 return (ext2_dirent *)((char*)p + le16_to_cpu(p->rec_len));
517 static inline unsigned
518 ext2_validate_entry(char *base, unsigned offset, unsigned mask)
520 ext2_dirent *de = (ext2_dirent*)(base + offset);
521 ext2_dirent *p = (ext2_dirent*)(base + (offset&mask));
522 while ((char*)p < (char*)de)
523 p = ext2_next_entry(p);
524 return (char *)p - base;
527 static inline void ext2_put_page(struct page *page)
530 page_cache_release(page);
533 static inline unsigned long dir_pages(struct inode *inode)
535 return (inode->i_size + CFS_PAGE_SIZE - 1) >> CFS_PAGE_SHIFT;
539 int ll_objects_destroy(struct ptlrpc_request *request, struct inode *dir);
540 struct inode *ll_iget(struct super_block *sb, ino_t hash,
541 struct lustre_md *lic);
542 int ll_mdc_cancel_unused(struct lustre_handle *, struct inode *, int flags,
544 int ll_mdc_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
545 void *data, int flag);
546 int ll_prepare_mdc_op_data(struct mdc_op_data *,
547 struct inode *i1, struct inode *i2,
548 const char *name, int namelen, int mode, void *data);
549 #ifndef HAVE_VFS_INTENT_PATCHES
550 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
553 void ll_pin_extent_cb(void *data);
554 int ll_page_removal_cb(void *data, int discard);
555 int ll_extent_lock_cancel_cb(struct ldlm_lock *lock, struct ldlm_lock_desc *new,
556 void *data, int flag);
557 int lookup_it_finish(struct ptlrpc_request *request, int offset,
558 struct lookup_intent *it, void *data);
559 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
562 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
563 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
564 int ll_writepage(struct page *page);
565 void ll_inode_fill_obdo(struct inode *inode, int cmd, struct obdo *oa);
566 int ll_ap_completion(void *data, int cmd, struct obdo *oa, int rc);
567 int llap_shrink_cache(struct ll_sb_info *sbi, int shrink_fraction);
568 extern struct cache_definition ll_cache_definition;
569 void ll_removepage(struct page *page);
570 int ll_readpage(struct file *file, struct page *page);
571 struct ll_async_page *llap_cast_private(struct page *page);
572 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
573 void ll_ra_accounting(struct ll_async_page *llap,struct address_space *mapping);
574 void ll_truncate(struct inode *inode);
575 int ll_file_punch(struct inode *, loff_t, int);
576 ssize_t ll_file_lockless_io(struct file *, const struct iovec *,
577 unsigned long, loff_t *, int, ssize_t);
578 void ll_clear_file_contended(struct inode*);
579 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
582 extern struct file_operations ll_file_operations;
583 extern struct file_operations ll_file_operations_flock;
584 extern struct file_operations ll_file_operations_noflock;
585 extern struct inode_operations ll_file_inode_operations;
586 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *);
587 extern int ll_have_md_lock(struct inode *inode, __u64 bits);
588 int ll_extent_lock(struct ll_file_data *, struct inode *,
589 struct lov_stripe_md *, int mode, ldlm_policy_data_t *,
590 struct lustre_handle *, int ast_flags);
591 int ll_extent_unlock(struct ll_file_data *, struct inode *,
592 struct lov_stripe_md *, int mode, struct lustre_handle *);
593 int ll_file_open(struct inode *inode, struct file *file);
594 int ll_file_release(struct inode *inode, struct file *file);
595 int ll_lsm_getattr(struct obd_export *, struct lov_stripe_md *, struct obdo *);
596 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
597 struct lov_stripe_md *lsm, lstat_t *st);
598 int ll_glimpse_size(struct inode *inode, int ast_flags);
599 int ll_local_open(struct file *file,
600 struct lookup_intent *it, struct ll_file_data *fd,
601 struct obd_client_handle *och);
602 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
603 int ll_mdc_close(struct obd_export *mdc_exp, struct inode *inode,
605 int ll_mdc_real_close(struct inode *inode, int flags);
606 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid, struct file
607 *file, size_t count, int rw);
608 #if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
609 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
610 struct lookup_intent *it, struct kstat *stat);
611 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
613 struct ll_file_data *ll_file_data_get(void);
614 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0))
615 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
617 int ll_inode_permission(struct inode *inode, int mask);
619 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
620 int flags, struct lov_user_md *lum,
622 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
623 struct lov_mds_md **lmm, int *lmm_size,
624 struct ptlrpc_request **request);
625 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
627 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
628 int *lmm_size, struct ptlrpc_request **request);
631 extern struct dentry_operations ll_init_d_ops;
632 extern struct dentry_operations ll_d_ops;
633 extern struct dentry_operations ll_fini_d_ops;
634 void ll_intent_drop_lock(struct lookup_intent *);
635 void ll_intent_release(struct lookup_intent *);
636 extern void ll_set_dd(struct dentry *de);
637 int ll_drop_dentry(struct dentry *dentry);
638 void ll_unhash_aliases(struct inode *);
639 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
640 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
641 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
642 int revalidate_it_finish(struct ptlrpc_request *request, int offset,
643 struct lookup_intent *it, struct dentry *de);
645 /* llite/llite_lib.c */
646 extern struct super_operations lustre_super_operations;
648 char *ll_read_opt(const char *opt, char *data);
649 void ll_lli_init(struct ll_inode_info *lli);
650 int ll_fill_super(struct super_block *sb);
651 void ll_put_super(struct super_block *sb);
652 void ll_kill_super(struct super_block *sb);
653 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
654 void ll_clear_inode(struct inode *inode);
655 int ll_setattr_raw(struct inode *inode, struct iattr *attr);
656 int ll_setattr(struct dentry *de, struct iattr *attr);
657 #ifndef HAVE_STATFS_DENTRY_PARAM
658 int ll_statfs(struct super_block *sb, struct kstatfs *sfs);
660 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
662 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
663 __u64 max_age, __u32 flags);
664 void ll_update_inode(struct inode *inode, struct lustre_md *md);
665 void ll_read_inode2(struct inode *inode, void *opaque);
666 int ll_iocontrol(struct inode *inode, struct file *file,
667 unsigned int cmd, unsigned long arg);
668 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
669 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
671 void ll_umount_begin(struct super_block *sb);
673 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
674 int ll_prep_inode(struct obd_export *exp, struct inode **inode,
675 struct ptlrpc_request *req, int offset, struct super_block *);
676 void lustre_dump_dentry(struct dentry *, int recur);
677 void lustre_dump_inode(struct inode *);
678 struct ll_async_page *llite_pglist_next_llap(struct ll_sb_info *sbi,
679 struct list_head *list);
680 int ll_obd_statfs(struct inode *inode, void *arg);
681 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
682 int ll_process_config(struct lustre_cfg *lcfg);
684 /* llite/llite_nfs.c */
685 extern struct export_operations lustre_export_operations;
686 __u32 get_uuid2int(const char *name, int len);
687 struct dentry *ll_fh_to_dentry(struct super_block *sb, __u32 *data, int len,
688 int fhtype, int parent);
689 int ll_dentry_to_fh(struct dentry *, __u32 *datap, int *lenp, int need_parent);
691 /* llite/special.c */
692 extern struct inode_operations ll_special_inode_operations;
693 extern struct file_operations ll_special_chr_inode_fops;
694 extern struct file_operations ll_special_chr_file_fops;
695 extern struct file_operations ll_special_blk_inode_fops;
696 extern struct file_operations ll_special_fifo_inode_fops;
697 extern struct file_operations ll_special_fifo_file_fops;
698 extern struct file_operations ll_special_sock_inode_fops;
700 /* llite/symlink.c */
701 extern struct inode_operations ll_fast_symlink_inode_operations;
703 /* llite/llite_close.c */
704 struct ll_close_queue {
706 struct list_head lcq_list;
707 wait_queue_head_t lcq_waitq;
708 struct completion lcq_comp;
711 #ifdef HAVE_CLOSE_THREAD
712 void llap_write_pending(struct inode *inode, struct ll_async_page *llap);
713 void llap_write_complete(struct inode *inode, struct ll_async_page *llap);
714 void ll_open_complete(struct inode *inode);
715 int ll_is_inode_dirty(struct inode *inode);
716 void ll_try_done_writing(struct inode *inode);
717 void ll_queue_done_writing(struct inode *inode);
719 static inline void llap_write_pending(struct inode *inode,
720 struct ll_async_page *llap) { return; };
721 static inline void llap_write_complete(struct inode *inode,
722 struct ll_async_page *llap) { return; };
723 static inline void ll_open_complete(struct inode *inode) { return; };
724 static inline int ll_is_inode_dirty(struct inode *inode) { return 0; };
725 static inline void ll_try_done_writing(struct inode *inode) { return; };
726 static inline void ll_queue_done_writing(struct inode *inode) { return; };
727 //static inline void ll_close_thread_shutdown(struct ll_close_queue *lcq) { return; };
728 //static inline int ll_close_thread_start(struct ll_close_queue **lcq_ret) { return 0; };
730 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
731 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
733 /* llite/llite_mmap.c */
734 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
735 typedef struct rb_root rb_root_t;
736 typedef struct rb_node rb_node_t;
739 struct ll_lock_tree_node;
740 struct ll_lock_tree {
742 struct list_head lt_locked_list;
743 struct ll_file_data *lt_fd;
746 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
747 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
748 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
749 __u64 end, ldlm_mode_t mode);
750 int ll_tree_lock(struct ll_lock_tree *tree,
751 struct ll_lock_tree_node *first_node,
752 const char *buf, size_t count, int ast_flags);
753 int ll_tree_lock_iov(struct ll_lock_tree *tree,
754 struct ll_lock_tree_node *first_node,
755 const struct iovec *iov, unsigned long nr_segs,
757 int ll_tree_unlock(struct ll_lock_tree *tree);
759 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
761 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
762 static inline __u64 ll_ts2u64(struct timespec *time)
764 __u64 t = time->tv_sec;
768 static inline __u64 ll_ts2u64(time_t *time)
774 /* don't need an addref as the sb_info should be holding one */
775 static inline struct obd_export *ll_s2obdexp(struct super_block *sb)
777 return ll_s2sbi(sb)->ll_osc_exp;
780 /* don't need an addref as the sb_info should be holding one */
781 static inline struct obd_export *ll_s2mdcexp(struct super_block *sb)
783 return ll_s2sbi(sb)->ll_mdc_exp;
786 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
788 struct obd_device *obd = sbi->ll_mdc_exp->exp_obd;
794 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
795 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
797 return ll_s2sbi(inode->i_sb);
800 static inline struct obd_export *ll_i2obdexp(struct inode *inode)
802 return ll_s2obdexp(inode->i_sb);
805 static inline struct obd_export *ll_i2mdcexp(struct inode *inode)
807 return ll_s2mdcexp(inode->i_sb);
810 static inline void ll_inode2fid(struct ll_fid *fid, struct inode *inode)
812 mdc_pack_fid(fid, inode->i_ino, inode->i_generation,
813 inode->i_mode & S_IFMT);
816 static inline int ll_mds_max_easize(struct super_block *sb)
818 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
821 static inline __u64 ll_file_maxbytes(struct inode *inode)
823 return ll_i2info(inode)->lli_maxbytes;
827 int ll_setxattr(struct dentry *dentry, const char *name,
828 const void *value, size_t size, int flags);
829 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
830 void *buffer, size_t size);
831 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
832 int ll_removexattr(struct dentry *dentry, const char *name);
836 #define LL_SA_RPC_MIN 2
837 #define LL_SA_RPC_DEF 32
838 #define LL_SA_RPC_MAX 8192
840 /* per inode struct, for dir only */
841 struct ll_statahead_info {
842 struct inode *sai_inode;
843 unsigned int sai_generation; /* generation for statahead */
844 atomic_t sai_refcount; /* when access this struct, hold
846 unsigned int sai_sent; /* stat requests sent count */
847 unsigned int sai_replied; /* stat requests which received
849 unsigned int sai_max; /* max ahead of lookup */
850 unsigned int sai_index; /* index of statahead entry */
851 unsigned int sai_hit; /* hit count */
852 unsigned int sai_miss; /* miss count:
853 * for "ls -al" case, it includes
854 * hidden dentry miss;
855 * for "ls -l" case, it does not
856 * include hidden dentry miss.
857 * "sai_miss_hidden" is used for
860 unsigned int sai_consecutive_miss; /* consecutive miss */
861 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
862 * is not a hidden one */
863 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
864 unsigned int sai_ls_all:1; /* "ls -al", do stat-ahead for
866 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
867 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
868 struct list_head sai_entries; /* stat-ahead entries */
871 int do_statahead_enter(struct inode *dir, struct dentry **dentry, int lookup);
872 void ll_statahead_exit(struct dentry *dentry, int result);
873 void ll_stop_statahead(struct inode *inode, void *key);
876 void ll_d_wakeup(struct dentry *dentry)
878 struct ll_dentry_data *lld = ll_d2d(dentry);
880 LASSERT(dentry->d_op != &ll_init_d_ops);
882 cfs_waitq_broadcast(&lld->lld_waitq);
886 int ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int lookup)
888 struct ll_sb_info *sbi = ll_i2sbi(dir);
889 struct ll_inode_info *lli = ll_i2info(dir);
890 struct ll_dentry_data *ldd = ll_d2d(*dentryp);
892 if (sbi->ll_sa_max == 0)
895 /* not the same process, don't statahead */
896 if (lli->lli_opendir_pid != cfs_curproc_pid())
900 * When "ls" a dentry, the system trigger more than once "revalidate" or
901 * "lookup", for "getattr", for "getxattr", and maybe for others.
902 * Under patchless client mode, the operation intent is not accurate,
903 * it maybe misguide the statahead thread. For example:
904 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
905 * have the same operation intent -- "IT_GETATTR".
906 * In fact, one dentry should has only one chance to interact with the
907 * statahead thread, otherwise the statahead windows will be confused.
908 * The solution is as following:
909 * Assign "lld_sa_generation" with "sai_generation" when a dentry
910 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
911 * will bypass interacting with statahead thread for checking:
912 * "lld_sa_generation == lli_sai->sai_generation"
914 if (ldd && lli->lli_sai &&
915 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
918 return do_statahead_enter(dir, dentryp, lookup);
921 /* llite ioctl register support rountine */
928 #define LLIOC_MAX_CMD 256
931 * Rules to write a callback function:
934 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
935 * returned to ll_iocontrol_register. Callback functions should use this
936 * data to check the potential collasion of ioctl cmd. If collasion is
937 * found, callback function should return LLIOC_CONT.
938 * @rcp: The result of ioctl command.
941 * If @magic matches the pointer returned by ll_iocontrol_data, the
942 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
944 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
945 struct file *file, unsigned int cmd, unsigned long arg,
946 void *magic, int *rcp);
948 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
949 unsigned int cmd, unsigned long arg, int *rcp);
951 /* export functions */
952 /* Register ioctl block dynamatically for a regular file.
954 * @cmd: the array of ioctl command set
955 * @count: number of commands in the @cmd
956 * @cb: callback function, it will be called if an ioctl command is found to
957 * belong to the command list @cmd.
960 * A magic pointer will be returned if success;
961 * otherwise, NULL will be returned.
963 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
964 void ll_iocontrol_unregister(void *magic);
968 #endif /* LLITE_INTERNAL_H */